Mistral Small 4
Mistral · Released 16 Mar 2026 · mistral-small-2603
$0.15
Input price / 1M tokens
#3 of 61, cheapest first
$0.60
Output price / 1M tokens
#4 of 61, cheapest first
256K
Context window
#59 of 71
183
Output tokens / second
#13 of 61
Summary
Mistral Small 4 is an open-weight model from Mistral, released on 16 Mar 2026.
At $0.15 input and $0.60 output per million tokens, it is #3 of 61 on input price, cheapest first.
Its 256K-token context window ranks #59 of 71.
Measured output speed is 183 tokens per second, #13 of 61.
Open-weight hybrid MoE model (119B total, about 6B active) that unifies instruct, reasoning (Magistral) and agentic coding (Devstral) with image input and configurable reasoning effort.
Benchmarks · 1 reported
| Benchmark | Score | Bar | Rank | Source |
|---|---|---|---|---|
| AA-LCR (Artificial Analysis Long Context Reasoning)with reasoning | 0.72% | – | Mistral launch postSelf-reported |
Self-reported means the lab ran the test itself. A rank appears only where several labs report the same version of a benchmark.
Sources
- API ID, version 26.03, date, license, context, price, parameters: docs.mistral.ai/models/mistral-small-4-0-26-03, checked 2 Oct 2026
- cached input price: docs.mistral.ai/inference/pricing, checked 2 Oct 2026
- batch discount, free plan API credits: mistral.ai/pricing, checked 2 Oct 2026
- launch post: image input, 256k context, architecture, AA-LCR score: mistral.ai/news/mistral-small-4, checked 2 Oct 2026
- output speed: artificialanalysis.ai/models/mistral-small-4, checked 2 Oct 2026
Building on Mistral Small 4?
Get the architecture right before the bill arrives.
We size caching, routing and fallbacks for your workload on a free call, and tell you where a cheaper model is good enough.